Exercising Growth Options through Seasoned Equity Offerings
Comparative Impact on the Stock Returns of Nordic issuers
*
Authors:
Andra-Maria Vasilescu Svitlana Malitska
Supervisor:
Ilan Cooper, Associate Professor, Ph.D.
Hand-in date: 30th of August, 2011 Campus:
BI Oslo
Examination code: GRA 19002 Master Thesis Programme: Master of Science in Financial Economics
This thesis is a part of the M.Sc. programme at BI Norwegian Business School. The school takes no responsibility for the methods used, results found and conclusions
drawn.
ABSTRACT
The apparent long-run abnormal underperformance of equity issuers has stirred great interest in finance until real (growth) options explanations have been successfully developed and tested on American SEO data in recent years.
Drawing on the existing literature and on a sample of Norwegian, Swedish and Danish seasoned equity offerings from 1997 to 2009 our paper highlights a risk pattern for the issuers around the SEO date consistent with the predictions of the real options theories, namely a risk run-up prior to issuance followed by a decrease in beta after issuance. We also find significant evidence of long-run abnormal performance for equally weighted SEO portfolios in factor regressions.
The magnitude and the significance of the intercepts are reduced or eliminated if the SEO portfolios returns are value-weighted instead of equally weighted. An investment factor long in low investment stocks and short in high investment stocks used to augment CAPM and Fama French regressions does not clear the underperformance but does contribute to a small reduction in the magnitude and significance of the intercepts in factor regressions. We believe that our mixed evidence generally speaks in favour of the growth options theories and calls for further research in this area.
Keywords: Seasoned equity offerings, Risk dynamics, Real investment
1 Introduction
Corporations are there to create value. Their capacity of value creation depends greatly on the investment decisions made by their managers. Such decisions involve the type and the timing of the investment undertaken, and the sources of financing. Our paper will explore the long-run impact of financing decisions on firm value, when choices entail equity financing, in particular through seasoned equity offerings. We are primarily interested in exploring the validity of the real options explanations for the observed patterns of systematic risk and of long-run stock performance of Norwegian, Swedish and Danish issuers.
The first section of our paper will briefly link our research topic to relevant financial theory (from the choice of capital structure to behavioral and real investment based explanations for the stock return pattern of seasoned equity issuers), and to previous empirical studies on the SEO risk and return profile. Our empirical analysis will start with a focus on the long-run risk dynamics of seasoned equity issuers before and after issuance. The focus will then shift on the hypothesis of long-run abnormal post-issuance negative performance of seasoned equity issuers, as well as on the real options explanations advanced in the academic literature for the observed SEO return pattern.
2 1. Literature review
1.1 Capital structure and external financing
The corporate financing activity is crucial to the well-functioning of a company and to the realization of its goals. Funds are used to finance growth and can come from either internal or operating sources such as retained earnings and supplier credit, or from capital markets (Figure 1) in the form of external financing.
Figure 1 The corporate financing activity
Any external corporate financing decision – be it debt or equity- impacts the capital structure of the company directly. In order to understand the causes and the effects of financing actions (e.g. equity offerings) on company performance, it is therefore useful to look at the determinants of the capital structure. There are three main theories trying to explain the choice of the debt-equity ratio by firms. These theories go beyond the original Modigliani-Miller (1958) capital structure irrelevance theorem, which states that in perfect markets the total value of the firm is indifferent to the choice of capital structure, although this choice does impact the way the pie is split between equity holders and debt claim holders.
Corporate financing activity
Internal
Retained
earnings Supplier credit Depreciation
External
Debt
Straight
Convertible
Bank loans
Securities offerings
Initial offering (IPO)
Seasoned Equity Offerings (SEO)
Private Equity, Venture Capital,
etc.
3 Trade-off theory
The first theory was developed by the same Modigliani and Miller through a correction to their original model (Modigliani and Miller 1963). The trade-off theory says that companies have optimal debt-equity ratios which can be estimated by weighing the benefits of debt (namely the deductibility of interest expenses for tax purposes contrasted with the non-deductibility of dividends) against its costs. The major costs of debt are caused by the probability of bankruptcy when the company incurs expenses related to the bankruptcy filing, as well as by instances of financial distress when defections by customers and suppliers are common.
The trade-off theory can thus be summarized by the following relationship:
Firm value=Vu + PV (tax shield) – PV (costs of financial distress)
The implication of the trade-off theory is that large, mature companies with limited investment opportunities should hold higher leverage ratios to take advantage of the tax deductibility of debt given low financial distress costs. By contrast, smaller companies with growth opportunities should avoid debt to preserve their capacity of chasing positive NPV projects (Graham and Harvey 2002).
Pecking order theory
An alternative perspective on the choice of capital structure is the pecking order theory which posits that actual corporate leverage ratios typically do not reflect capital structure targets. This theory is related to the widely observed corporate practice of financing new investments using internal funds with priority, and only when these are depleted through external financing in the form of debt and then of equity offerings, in this order of preference (Myers 1984). The pecking-order theory sees equity offerings as the most expensive form of external financing because of information asymmetries between investors and managers. The issue of information asymmetry will be expanded in the following sections.
4 The free cash flow theory
The third theory around the choice of capital structure focuses on the agency costs associated with the free cash flow available to managers after they undertake positive NPV projects (Jensen 1986). Namely, payouts to shareholders in the form of dividends reduce the financial resources available to managers for investments.
Therefore managers have an incentive to retain earnings and to grow their firms sub-optimally. Growth means more resources under managers’ control and higher management compensation. Furthermore, internal financing avoids the issue of active monitoring by capital markets of managers’ activity when external financing is used. Jensen shows that debt can improve organizational efficiency.
First, interest payments curb overinvestment by reducing the amount of free cash flow at managers’ discretion. Second, the consequences of a failure to make debt service payments motivate the managers to run their organization more efficiently.
In line with the free cash flow theory, Jensen reiterated what Smith (1986) had empirically revealed, namely that most leverage-increasing transactions like stock repurchases and exchange of debt for stock are followed by increases in stock prices. Conversely, most leverage-decreasing transactions like equity issues or exchange of common stock for debt lead to significant falls in stock price as the market penalizes instances where managers have more resources under their control.
1.2 Seasoned equity offerings
The term SEO is employed in the academic literature to refer to equity offerings performed by firms which are already publicly listed. According to Ritter (2003), practitioners refer to such transactions with the term “follow-on offerings”.
Although sometimes SEOs are also referred to as secondary offerings, it is important to distinguish them from a secondary offering in the sense of a transaction where shares are sold by existing shareholders, as contrasted to a primary offering where shares are sold by the company.
Although they obey a concise definition, seasoned equity offerings are complex financial transactions, differentiable through key elements of their engineering design such as: the targeted investors (e.g. the public at large or existing
5 shareholders), the market (domestic issues or global issues), the type of proceeds (cash or equity), the flotation method or the marketing and selling mechanism (e.g. private placements, firm commitment, best efforts, rights, standby rights, auctions), just to name a few. Eckbo, Masulis and Norli (2007) provide a comprehensive list of SEO flotation methods.
As evidenced in the literature, certain features of SEOs are more typical of certain markets and display a changing pattern. In most countries, SEOs by public firms are typically conducted as rights offerings, whereas very few SEOs are conducted as public offerings. Thus, if rights are not used, the firm can attempt to sell the issue directly to the market with no financial intermediary, place the issue with a private group of investors (a private placement), or employ an intermediary, usually an investment banker or underwriting syndicate. In addition, stock can be sold through the issuance of convertible securities, warrants and stock options and through the establishment of dividend reinvestment, employee stock ownership and management compensation plans (Eckbo and Masulis 1995). Ekbo (2008) talks about the disappearing rights offer phenomenon in the USA. Rights offers, quite popular between 1935 and 1955, usually occur because the typical company charter stipulates a preemptive right in favor of existing shareholders to any new tranches of equity intended for sale. Such rights take the form of temporary warrants to purchase the new stock on a pro-rata basis based on existing holdings and at a discount relative to the prevailing market price. Eckbo (2008) believes that the underlying cause behind this phenomenon is that the cost of such rights may be prohibitively high in large companies with fragmented ownership.
By contrast, rights offers have remained popular in Europe and Asia until recently when there as well, companies have grown bigger in size and increased market participation has led to disperse ownership. Equity rights offerings are still popular in Greece for example (Cohen, Papadaki, and Siougle 2007). The pre- emptive right of first refusal is a long tradition in the UK and among the listing requirements on London Stock Exchange. It has also been stipulated in the European Community's Second Company Law Directive (1977) and, since 1980, in the UK Companies Act (Armitage 1998). However, Japan has experienced a trend away from rights offerings after the mid-1990s (Eckbo, Masulis, and Norli 2007). A similar tendency has emerged in French SEOs and on Oslo Stock
6 Exchange (Bøhren, Eckbo, and Michalsen 1997). Bøhren et al. agree that after 1993, following increased share ownership by domestic and foreign investors of the stocks listed on Oslo Stock Exchange, issuers began to switch to standby offers as the overriding flotation method.
Reasons for raising capital through SEOs
The most common reason why companies raise capital is to finance growth through real investment (e.g. capital expenditures, raw materials, new positive NPV projects). Eckbo, Masulis and Norli (2007) provide a comprehensive survey of other reasons explored in the literature: to change capital structure, to exploit private information about the intrinsic value of securities, to finance mergers and acquisitions, to facilitate asset restructuring such as spin-offs, to improve the liquidity of existing shares, to shift wealth and risk bearing among classes of securities, and privatizations. In a survey of American CFOs, Graham and Harvey (2002) indicate the following motives for common stock offerings in this order of prevalence: earnings per share dilution, perceived equity undervaluation/
overvaluation, recent stock price run-ups, providing shares to employee stock option plans, maintain target debt/equity ratios, diluting holdings of certain shareholders, “stock is our least risky source of funds”, holding similar amount of equity as same-industry firms, favorable investor impression versus debt issuance, no other sources of funds available, “stock is the cheapest source of funds”.
The underlying factors behind the decision to issue securities come from several core areas of finance like: capital structure, managerial investment incentives, contract theory, and asset pricing. Thus it should come as no surprise that there is no consensus in the literature on the economic implications of the equity issuance decision at company level.
1.3 The announcement effect versus long run stock performance
There is extensive empirical evidence in the finance literature that stock returns are impacted by external financing events in the short and in the long run.
7 1.3.1 The SEO announcement effect
In the short-run, many academics talk about a negative 2-day stock price reaction associated with security offering announcements, called the announcement effect (Smith 1986). Most studies performed on the US stock markets reveal a 2-day post-announcement abnormal return in the range of -1.0 % and -3.0% e.g. Asquith and Mullins (1986), Bayless and Chaplinksy (1996).
Information asymmetry
The leading explanation of this negative reaction is information asymmetry and was introduced by Myers and Majluf (1984). This explanation is pervasive in the literature under different other but equivalent labels: adverse selection, overvaluation etc. In corporate governance terms, the management acts as principals of the shareholders and their main objective is to increase shareholders’
wealth. Strong-form market inefficiency is assumed such that at any point in time, the management has superior information which tells them whether the stock is undervalued or overvalued. Given company’s objective, equity will be issued when the management has information that the stock is overvalued and debt issuance is not preferable. The market, knowing this rule of action, would penalize what they believe to be a signal of overvaluation.
Information signal about the investment policy
However, the announcement is also a signal about the investment policy of the company. If the market believes that the company will use the proceeds to engage in profitable projects then the stock price may increase. Conversely, if the market suspects that the management is squandering corporate resources, a price decrease will follow the issuance. Thus the information asymmetry proposition can be complemented by the free cash flow theory introduced earlier in this section.
Indeed Ritter (2003) shows that the additional equity resources raised through SEOs are relaxing the existing constraints on management’s proclivity for
“empire-building” or excessive growth. According to the free cash flow theory, such a constraint is the existing debt level, diluted through equity issuance. This implies that agency conflicts between shareholders and managers are intensified.
8 Other hypothetical explanations for the SEO negative announcement effect have been advanced in the literature. The following explanations were compiled by Smith (1986): optimal capital structure, implied cash flow changes (stock price changes reflect information about expected changes in net operating cash flows), unanticipated announcement (stock price changes reflect only the unanticipated component of the offering announcement and therefore their magnitude will vary inversely with the degree of announcement predictability ceteris paribus);
ownership changes (changes in the structure of control rights in the firm affect the value of firm’s equity). Armitage(1998) highlights the price inelasticity of demand for new shares. The announcement effect has also been interpreted as an indirect flotation cost (Eckbo, Masulis, and Norli 2007).
1.3.2. Evidence of long-run abnormal performance
Recent years have seen a surge in the number of studies on the topic of SEO long- run performance, mostly triggered by the emergence in the 1990s of the comprehensive, easy-to-use database of new corporate issues provided by the Security Data Company (SDC), a part of Thomson Reuters information service (Eckbo, Masulis, and Norli 2007).
However, the evaluation of issuers’ performance around SEOs remains a controversial issue despite the abundance of studies carried out over the past two decades. Bayless and Jay (2007) provide a short review of the recent work in this field. It all seems to have started with a potential stock market anomaly coined as
“the new issues puzzle” by Loughran and Ritter (1995). They show that during five years following the offering, US companies that issued equity between 1970 and 1990 either through an IPO or an SEO significantly underperformed relative to non-issuers matched by size, book-to-market and other firm characteristics.
Therefore, the evidence on the long-run performance of firms conducting SEOs is that issuing firms have relatively low returns compared to non-issuers or a benchmark in the 3–5 years after the SEO.
1.3.2.1 Methodology: Buy-and-hold returns versus factor regressions Most of the empirical studies on the long-run performance of SEOs have employed two methodologies typical for studies of long-run abnormal stock
9 performance: buy-and-hold returns and factor regressions (Lyon, Barber, and Tsai 1999).
The buy-and-hold methodology consists of matching the issuers to non-issuers based on a single or on multiple characteristics such as size, and book-to-market and then calculating the t-statistics using annual holding-period returns of issuing firms relative to matched non-issuing firms (Loughran and Ritter 1995) or relative to portfolios of the latter (Lyandres, Sun, and Zhang 2008). The weakness of the buy-and-hold approach is that unbiased t-statistics are difficult to obtain due to three main biases: the skewness bias – the fact that long-horizon abnormal returns are positively skewed, the survivor bias- affecting the sample of non-issuing matching firms, and the rebalancing bias – differences in compounded returns between issuers and non-issuers resulting from rebalancing techniques (Lyon, Barber, and Tsai 1999). Lyandres et al. (2008) apply skewness-adjusted t- astatistics as a correction. Yet, factor regressions avoid the bias issue altogether and have been preferred by many researchers.
These models have been inspired by the seminal 3-factor model developed by Fama and French (1993) :
where is the excess return on a portfolio in period t, is the realized market risk premium in period t, is the return on a portfolio of small stocks minus the return on a portfolio of big stocks in period t, and is the return on a portfolio of value stocks (i.e. with high book-to-market ratios) minus the return on a portfolio of growth stocks (i.e. with low book-to-market ratios) in period t. The non-zero intercepts of this regression are interpreted as (significant or non-significant) abnormal performance.
The information contained in the following table represents a collection of empirical results from American SEO studies which have employed the two methodologies. A part of this information has been compiled by Ritter (2003).
10 Table 1 - Evidence on long-run abnormal performance of SEOs in the USA
This table reports the summary results of long-term seasoned offerings underperformance studies.
Panel A is based on the buy-and-hold (BHR) methodology. The mean buy-and hold returns are represented for SEOs and their matches (based on size and book-to-market). Values in brackets indicate the t-statistics. The information contained in this table was compiled by Ritter (2003) except the study by Lyandres et al.
Panel A: Buy-and-hold abnormal returns (BHR) Studies Sample
size Period Horizon Mean BHR Annualized
diff.
SEOs Match Diff (Mitchell
and Stafford
2000) 4439 1961-1993 3 years 34.8% 45.0% -10.4% -2.7%
(Eckbo, Masulis, and
Norli 2000) 3315 1964-1995 5 years 44.3% 67.5% -23.2% -4.8%
(Jegadeesh
2000) 2992 1970-1993 5 years 59.4% 93.6% -34.2% -4.9%
(Lyandres, Sun, and
Zhang 2008) 10084 1970-2005 5 years NA NA -50 % NA Panel B: Fama French 3-factor regression
Studies Sample
size Period Equally weighted
intercepts Value-weighted intercepts (Eckbo, Masulis, and
Norli 2000)1 1704 1964-1997 -0.12 (-0.65) -0.17 (-1.12) (Mitchell and Stafford
2000)2 4911 1961-1993 -0.33 (-5.19) -0.03 (-0.44)
(Jegadeesh 2000)3 2992 1975-1995 -0.45 (-5.07) -0.33 (-2.84)
(Loughran and Ritter
2000)4 6461 1973-1996 -0.47 (-5.42) -0.32 (-3.00)
(Lyandres, Sun, and
Zhang 2008)5 10084 1970-2005 -0.39 (-3.52) -0.35 (-3.04)
1 Amex/NASDAQ, excluding utilities
2 Incl. utilities; an SEO firm is kept in the portfolio 5 years after issuance;
use monthly returns
3An SEO firm is kept in the portfolio 5 years after issuance
4 Excl. utilities; an SEO firm is kept in the portfolio 3 years after issuance
5 Incl. utilities; an SEO firm is kept in the portfolio 3 years after issuance
The academic reactions to this empirical evidence have come in two main forms.
First, Brav, Gecy and Gompers (2000) conclude that the SEO long-run underperformance is more related to the characteristics of the issuing firms than to the actual issuance decision. Thus, in a sample of SEOs from 1975 to 1992, they find that underperformance is concentrated in small issuing firms with low book- to-market ratios such that “the stock returns following equity issues reflect a more
11 pervasive return pattern in the broader set of publicly traded companies”. Second, researchers have developed and tested several plausible hypotheses in an attempt to explain this apparent anomaly.
1.3.2.2 The “bad” model problem
One hypothesis is the “bad model” problem. Market efficiency requires that, on average, there should be no abnormal returns after an event if an appropriate benchmark is used. As highlighted by Loughran and Ritter (2000), the problem is that tests of market efficiency are always joint tests of a (theoretically supported) model of market equilibrium and of the existence of abnormal returns. But in the buy-and-hold methodology, matching issuers to non-issuers on size and book-to- market is supported empirically, rather than theoretically, therefore the abnormal returns reported in Table 1 cannot be considered enough evidence for or against market efficiency. At the same time, it is doubtful that the relatively low post- issue returns of the issuers can be connected to a lower level of risk, since as Ritter (2003) shows, issuing firms are highly exposed to systematic risk according to the Fama-French coefficients.
The “bad model” hypothesis has thus encouraged researchers to explore potential non-priced risk premia or risk patterns related to external financing events, ignored by existing models. For example, Eckbo, Masulis and Norli (2000) have shown that forming zero cost portfolios short in issuing stocks and long in matched non-issuing stocks yields statistically insignificant abnormal returns when a specific six factor regression is employed. Eckbo et al. empirically chose six macroeconomic factors such as: the value-weighted CRSP market index, the return spread between long and short maturity Treasury bonds, the return spread between long and short maturity T-bills, and the unexpected inflation (inflation shocks). They have argued that the liquidity premium on SEOs is low since the increased amount of outstanding shares makes them more liquid. Overall, the equity seems to carry less risk after the SEO event, which explains the post- issuance lower returns.
12 1.3.2.3 Financial leverage
Following Hamada (1972), many researchers have examined whether risk changes discretely at the time of equity or debt offerings due to changes in financial leverage. Such studies have consistently unveiled a positive correlation between the sign of the financial leverage changes and the sign of the impact on stock prices (e.g. (Asquith and Mullins 1986). Eckbo, Masulis and Norli (2000) argue that the decrease in leverage induced by an equity offering reduces the level of systematic risk exposure of the issuers.
1.3.2.4 Behavioral biases
Other important hypotheses relate to behavioral biases like market timing (Cohen, Papadaki, and Siougle 2007) and investor overconfidence over the precision of private information (Daniel, Hirshleifer, and Subrahmanyam 1998). The survey compiled by Graham and Harvey (2002) present evidence that the decisions to issue equity by US corporate executives are heavily influenced by behavioral biases.
1.3.2.5 Real investment and growth options
Another interesting hypothesis was crystallized through the research carried out in the area of optimal investment and production-based asset pricing. When asset prices are used to explain investment growth, academia talks about the q-theory of investment. When investment growth is used to explain asset prices, then we deal with production-based asset pricing (Porter 2005).
First, Cochrane (1991, 1996) introduced a theoretical production-based asset pricing model similar to the consumption-based model. A consumption-based asset pricing model relates asset returns to the marginal rate of substitution through an optimization of consumer’s utility function. Similarly, a production- based asset-pricing model relates asset returns to the marginal rate of transformation through a production function which gives the producers’ first order conditions for the optimal inter-temporal investment demand. The key concept in production based asset pricing is investment return (not to be confused with ROI) which represents the marginal rate of return which firms earn by
13 deviating from the optimal investment level through time, such that the deviations cancel each other and the production plan remains unchanged.
Cochrane (1991, 1996) has extended the q-theory of investment initiated by Tobin (1969) to reveal a negative relationship between real investment and expected stock returns. The ratio of an asset’s market value over the replacement cost of the same asset has been labeled Tobin’s q. For individual companies, Tobin’s q can be approximated as the ratio of the market value of equity over the book-value of assets. If a company is fairly evaluated by the market, then its q should be equal to 1.0. A q below or above unity suggests that the company is under or overvalued respectively. Alternatively, a q greater than 1.0 suggests that the market value reflects some assets which are not recorded in the balance sheet of the company.
These may be intangible assets such as growth options. Tobin’s marginal q is the ratio of the market value of an additional unit of capital to its replacement cost.
Cochrane shows that firms invest more when their marginal q is high, and that a high marginal q is associated with a low cost of capital.
Cooper and Priestley (2011) show that systematic risk falls during large investment periods in accordance with the q-theory of investment and the returns of a factor formed on investment-to-assets help forecast aggregate economic activity.
According to Lyandres et al. (2008), real investment is an important driving force behind the “new issues puzzle” because of the negative relationship between real investment and expected returns. The central finding of Lyndres et al. is that a new investment factor, long in low investment-to-assets stocks1 and short in high investment-to-assets stocks, explains a substantial part of the previously reported abnormal performance in the case of new issues such as IPOs, SEOs and convertible debt offerings. This factor is used to extend the Fama and French (1993) three-factor model. Lyandres’ et al. investment factor earns a significant
1The investment-to-assets ratio has been measured as the annual changes in gross property, plant, and equipment plus the annual changes in inventories divided by the lagged book value of assets.
14 average return. In addition, firms that issue equity and convertible debt appear to invest much more than matching non-issuers. Lyandres et al. conclude that adding the investment factor into standard factor regressions explains on average about 75% of the SEO underperformance.
Berk, Green and Naik (1999) are among the first academics to exploit the concept of real options as a possible link between real investment and the stock return dynamics of SEO firms. In their model, firms own two kinds of assets: assets that are in place and currently producing cash flows, and options to make positive NPV investments in the future. The projects carrying lower systematic risk are the most attractive to the firm and they subsequently lead to an increase in firm value.
At the same time, the overall level of systematic risk of the firm will diminish as a result of such investments, and the firm will experience lower returns in the future.
Carlson, Fisher and Giammarino (2006) develop a theoretical model of risk dynamics around an SEO using real options. Their model assumes an all-equity firm and does not rely on changes in financial leverage. The intuition behind this framework is that real investment transforms risky expansion options into less risky assets in place. This is why the riskiness of a company (as measured by its market beta for example) should decrease after the event, if the proceeds are used to finance real investment. This intuition challenges the traditional view that increases in capital expenditures have to be accompanied by positive stock price reactions (Trueman 1986) since they signal the availability of positive NPV projects (Jensen and Meckling 1976).
De Andres et al. (2008) infer that a firm’s beta is the weighted average of the betas of its assets-in-place and of its growth options:
where and represent, respectively, the beta and the total value of the firm i;
and measure, respectively, the beta and the value of its assets-in-place;
and measure, respectively, the systematic risk and the value of its growth options.
15 Carlson, Fisher and Giammarino (2010) serve as a key reference for the purpose of our paper, as they show that market betas of the equity issuing firms run up prior to the seasoned equity issuance and decline thereafter suggesting a similar pattern in the systematic risk of the issuers around the SEO events.
Our empirical approach will focus on the real-options based theory of the long- run stock returns dynamics around SEOs, given the relative novelty of this theoretical framework, as well as its capacity to arguably clear the new issues puzzle, hitherto considered an anomaly in finance.
Since most of the empirical studies testing the real options hypothesis focus on the stock market in the US, we will test the external validity of these studies in the case of the Nordic stock markets: Norway, Sweden and Denmark.
We will also investigate the long-run performance of SEOs following the investment factor methodology suggested by Lyandres et al. (2008).
2. Development of hypotheses
H1: Systematic risk increases before the SEO date and decreases thereafter.
One of the objectives of our paper is to explore the equity risk dynamics of issuers around seasoned equity offerings. If the conclusions reached by Carlson, Fisher and Giammarino (2010) are viable, the average market betas of our seasoned equity issuers should increase prior to issuance and decrease thereafter, in line with the predictions of the real options theories.
H2: Systematic risk dynamics around an SEO is more significantly impacted by the exercise of growth options than by the change in leverage.
While the results displayed by Carlson, Fisher and Giammarino (2010) strongly support a real-options based explanation of the risk dynamics around SEO events, we believe it would be interesting to explore to what (differing) degrees changes in betas are driven by increases in real investment and by changes in financial leverage respectively. The basic intuition is that exercising a growth option should induce a lower post-SEO beta. At the same time, increased equity financing
16 deleverages a company and should also contribute to a decrease in beta. Figure 2 provides a summary of these influences.
Figure 2 Changes in company’s risk profile induced by changes in financial leverage and by exercising growth options
Factors affecting systematic
risk Effects on systematic risk (Beta) Decrease in financial leverage
Exercise of growth options Resulting effect
These unidirectional effects on beta question the findings of Carlson et al. (2010), which empirically attribute the risk dynamics of the SEOs solely to realizations of real options.
H3: SEO firms display significant negative long-run abnormal performance.
Inspired by existing literature on SEO long-run performance, another major objective of our paper is to test the hypothesis of long-run abnormal negative performance of SEO firms. We will investigate whether issuers generate significantly negative abnormal returns in the long-run, using factor regressions (i.e. the CAPM and the Fama French three factor model).
H4: An investment factor long in low investment stocks and short in high investment stocks reduces the abnormal performance of the SEO portfolio.
If we find evidence of underperformance, we will investigate the real investment- based explanation of underperformance by augmenting standard factor regressions with an investment factor following the methodology proposed by Lyandres et al.
(2008).
17 3. Empirical implementation
3. 1 Risk dynamics around issuance events
In this section of our paper we will investigate the average beta behavior of our sample firms around the issuance dates, using the event study methodology, with an eye to the approach adopted by Carlson, Fisher and Giammarino (2006). We will estimate the average equally-weighted beta of our full samples of issuers, as well as the betas of sub-samples of stocks. In particular, we are interested in forming two subsamples: country subsamples and R&D intensive companies. Furthermore, we will look at the dynamics of the value-weighted average beta of our sample of issuers to check if and how size influences issuers’ risk pattern.
3.1.1 Data description
This analysis is based on three samples of Norwegian, Swedish and Danish seasoned equity offerings. The data samples are described in Table 1 from the Appendix. Certain constraints were employed for our research purposes. First, the samples are drawn from similar time periods for the three markets (Norway 1997- 2005; Sweden 1997-2005; Denmark 2000-2005) and we did not sample any SEOs after 2005 in order to avoid the patterns of stock behavior generated by the financial crisis (an extreme event).
Secondly, following Carlson et al. we tried to limit the impact of the issue event to the center of the time window of five years. Thus for each company in the list we checked for absence of additional stock issues two years before and three years after the issue of interest.
Finally, the SEO sample includes public and private equity placements and will exclude employee stock offerings (which are not primarily meant to raise capital for investments), as well as financial institutions.
Table 1 reports data characteristics analyzed from several perspectives. Our final sample consists of 186 issues performed by 177 companies in Norway (78 SEOs), Sweden (83) and Denmark (25). Both for issue size and the fraction of issue to the market value, we can observe a substantial dispersion between the average and
18 median due to outliers in the upper side. By analyzing medians as a more robust measure, we note similar sizes for Norwegian and Swedish issues at around 85 million NOK with somewhat lower sizes for Danish issues. From a historical perspective it is interesting to note that the values of issue size intuitively follow the pattern of market indexes, falling in 2002-2003, and rising afterwards. It might be also interesting to look at the industry distribution of the companies included in the sample, though just as for the other characteristics, we should abstain from making more general inferences about patterns in the absolute and relative sizes of the issues, due to the small number of companies sampled from each industry/year.
For the purpose of this section, we have extracted the following information from Datastream (Thomson Reuters):
Daily returns for the SEO firms in each country (RI);
Daily returns for market indexes: OSLO EXCHANGE ALL SHARE – OSLOASH (RI), OMX Stockholm 30 (OMXS30) – (DSRI-Datastream calculated Total Return Index); OMX Copenhagen (OMXC20) - (DSRI-Datastream calculated Total Return Index);
Accounting information: Research and Development-to-sales (datatype WC08341).
For consistency, throughout the paper we have used the Total Return Index datatype (DS Menmonic: RI) to compute returns. This index shows the theoretical variations in the value of a stock, assuming that dividends are reinvested and adjusting for stock splits and repurchases. For a detailed description of the RI datatype, please refer to the Note on the Total Return Index in the Appendix.
3.1.2. Methodology
In estimating the average betas of our SEO samples we followed the approach of Carlson et.al (2010) supplemented by the robustness methodology first established by Dimson (1979). We have looked at the long-term beta dynamics of issuing firms, 2 years before issuance and 3 years after issuance. The length of the time window is consistent with Lyandres’ choice (2008). By contrast Carlson et al.
19 (2006) use a five-year long pre- and post-issuance event window, but has shown that most of the issuance effect on stock returns occurs within 2 years before and 3 years after the event.
In order to obtain the beta time-series in the first place we followed two separate steps: first estimated betas for the uniform period for whole samples, and afterwards synchronized beta series for each company in accordance to issue date.
Beta estimation for the whole period
For each eligible company we extracted the total return index (RI) as well as the corresponding market RI for the period 01/01/1993 – 27/05/20112. Worth noting, for companies with several stock classes listed, we chose only the major security where the share class was not specified on the issue list.
Beta, or the slope of the regression line linking stock returns to the market return was estimated by employing matrix operations formulas of the form:
. Here X corresponds to the natural logarithm of the daily market return, and y to the natural logarithm of the daily (issuer’s) stock return. More specifically, for each issuing company at every daily point we estimated the beta over a certain previous period (using estimation windows of 1 month, half a year, and one year before the beta estimation point) by rolling over the estimation window day by day. This technique enabled us to obtain daily estimates of monthly, semiannual and annual betas, and therefore a dataset with more frequent beta estimates than those included in the dataset reported by Carlson et al. (2010).
Synchronizing and averaging betas
The date of the SEO event is the focus of our risk dynamics estimation. In order to obtain the average beta dynamics for all the issuers around the (general) SEO event date, we have synchronized the beta series of every firm so that the issuance date is placed at day “0”. The 520 daily beta estimates preceding the SEO date
2 Since the latest issues considered in this section happen in 2005, our analysis requires a far
shorter risk estimation period – up to 2009. Thus ending date as of 27/05/2011 can be considered somewhat arbitrary.
20 (two years before issuance3) and the 780 daily estimates following the SEO date (three years after issuance) are placed accordingly before and after day “0” on the timeline (see Figure 1 in the Appendix for an illustration of this procedure). We have implemented this procedure with monthly, semiannual and annual beta estimation windows.
Finally we average the synchronized beta estimates across all companies. The resulting average beta time series provides the basis for the graphical illustration of the risk dynamics around the SEO. We have illustrated the equally weighted and the value weighted risk dynamics of our SEO sample in Figure 2 and Figure 3 from the Appendix.
3.1.3 Robustness check motivation and methodology
The risk measurement formula previously described is widely accepted and used.
However, despite the apparent advantages, employing frequent but unstable daily returns data can hide certain pitfalls on the way to unbiased and consistent beta estimation. First noted by Fama (1965) and Fisher (1966) and further investigated by Scholes and Williams (1977) and Dimson (1979), biased OLS estimator of market beta is a significant feature of thinly traded securities. Not surprisingly, the returns data of Nordic issuers, as well as other small markets, does exhibit non- synchronous nature (market is dominated by securities that are not traded every day). Explaining the cause of Finnish stock market serial correlation Berglund et al. (1988; 1989) refer to thin trading as one of the major reasons. Subsequent works illustrate the importance of controlling for non-synchronous trading when measuring risk. Similar studies were performed by Bartholdy and Riding (1994) with data from New Zealand.
The principle behind the non-synchronous data problem is the following. A standard market model predicts true returns for security j in period t to be the function of market returns in the same period4.
3 We used the convention of 260-trading day in a year mainly based on the actual count of trading
days from DATASTREAM. Thus our monthly, semiannual and annual estimation windows consist of 21 days, 130 days, and 260 days respectively.
4 The following analytical illustration is largely based on (Cohen et al. 1983)
21
(1)
Observed returns, in contrast, have a stochastic nature to a large extent and are a function of true returns.
(2)
where for is a random variable that comprises a delay distribution.
Thus creates the delayed impact of the return generated in period t on the actually observed returns in the time window . Since the structure of
differs across securities, the observed returns for each of them will adjust asynchronously to their aggregate index, namely true market returns. Cohen et al.
(1983) show that with such asynchronous adjustments, cross-serial correlation is introduced into observed returns, and observed beta estimates are biased.
Technically speaking, the major source of this econometric problem stems from the covariation between the regressor and the residual . As for the resulting betas estimates, they tend to be biased downwards for infrequently traded stocks, and upwards for frequently traded ones.
The academic literature presents a set of alternative bias-correcting techniques, mostly derived from the basic studies of Scholes and Williams (1977) and Dimson (1979). The Scholes-Williams procedure requires estimating single-factor regressions in the simple market model form:
(3)
Consistent beta is subsequently calculated as
(4) where , , and represent lag, contemporaneous and lead market slope measures, while r is the first-order, serial correlation coefficient for the market index.
On the other hand, Dimson employed the multiple regressions in the form
22
(5)
And the Dimson’t beta is obtained by summing the slope estimates
(6)
Subsequent analytical research by Fowler and Rorke (1983) proved the inconsistency of Dimson’s technique and lead to the development of a correcting procedure incorporating both Scholes and Williams’ and Dimson’s frameworks.
Besides the theoretical generalization, the following beta estimation formula is also adapted for working with large amounts of time-series in Excel by incorporating only single regression estimates. Thus our risk dynamics analysis was performed through the following model:
(7)
where , , , , are the OLS regression estimators of , , - , , respectively, N is the number of leads and lags,
- is the observed security beta,
- is the observed intertemporal lag5 market beta,
- is the observed intertemporal lead market beta,
- is the observed intertemporal lag security beta,
- is the observed intertemporal lead security beta,
- is the true security beta, - is the observed daily return of company j in period t ,
5 Here the lead and lag definitions should be interpreted in the sense used by Scholes and Williams (1977) who view them from company’s perspective, while Cohen et al.(1983) take the opposite view, referring to the leads and lags of the market return. This small ambiguity makes no difference for the calculations.
23 - is the observed daily return of market index in period t,
-N is the number of leads and lags considered necessary to capture the delays in company returns reactions.
Partly following the analysis of Carlson, Fisher and Giammarino (2010) our paper describes the lead and lag structure of up to 2 and 5 leads and lags, in addition to the contemporaneous beta dynamics.
As a result, for each company we obtain 11 daily time series of beta estimates – starting with regressing firm returns on the 5-day market lead (firm lag) and ending with 5-day market lag (firm lead).
For each of the three markets we perform similar calculations to obtain 11 time series of market autocorrelation estimates at each daily point in time. Using formula (7) for each issuer we obtain the series of adjusted sum betas. Since our robustness check involves comparison of the series of contemporaneous, in addition to 2 and 5 leads and lags, formula (7) and the subsequent adjusted sum betas should be adjusted by the number of terms included. For example for contemporaneous beta a simple regression of firm returns on the market returns is needed to obtain the slope coefficient. For 5 leads and lags series, we would add up slope coefficients of regressing firm returns in certain period on the 5, 4, 3, 2, and 1-day market lead, the contemporaneous returns and the 1, 2, 3, 4, 5-day market lag (firm lead). After obtaining the sum we just divide it by the sum of the similar estimates of market autocorrelation. Again, all the basis and final series start from 01/01/1993 and end on 27/05/2011 for all countries.
The syncronization of the beta series has been implemented in the same way as illustrated in Section 3.1.2 for the contemporaneous beta.
Table 2 in the Appendix shows that the upward sloping beta dynamics prior to issuance and the downward sloping beta dynamics post-issuance, plotted using contemporaneous market returns is robust to adjustments for asynchronous trading using the methodology suggested by Fowler and Rorke (1983).
24 3.1.4 Results and interpretations
We will first introduce the empirical results for the aggregate sample of issuers then we will present the results of the same analysis performed on specific sub- samples in order to better understand how growth options are likely to influence the risk dynamics of the issuers.
3.1.4.1 Aggregate sample of issuers
The graph below illustrates how the average market beta evolves through time. In tune with the American SEO firms (Carlson, Fisher, and Giammarino 2010), Nordic issuers are also characterized by increasing risk several months before the SEO event and a smooth decrease in risk several months thereafter. Including up to 5 leads and lags to the adjusted sum beta (the upper middle-dark line) makes this trend even more pronounced, and the beta value more logical. It is indeed intuitively appealing to assume that the true average beta of the sample should be closer to 1.0, the beta of the market portfolio.
0 0,2 0,4 0,6 0,8 1 1,2 1,4 1,6
-520 -390 -260 -130 0 130 260 390 520 650 780
βeta
Figure 3: Equally-weighted Semiannual betas - Nordic issues (186)
Contemporaneous beta
Sum of up to 2 lead and lag betas Sum of up to 5 lead and lag betas
Days before (negative values) and after (positive values) issuance
25 A quantitative description of the results displayed graphically is provided in Table 2 in the Appendix. In particular, we have tried to reflect the change in beta over time by taking the difference of beta estimates between the pre-issuance and the post-issuance period. We can see that the difference between the beta estimated on the day of the issue and the beta estimated 1 and 2 years before the issue, regardless of the estimation window (monthly, semiannual (with different adjustment for leads and lags), or annual) is mostly insignificant for all subsamples, including the Nordic aggregate. However, the differences between the 1, 2 and 3-year post issuance betas and the beta estimates on the issuance date are negative and significantly different from zero. This means that there is indeed a significant decrease in the average beta 2 and 3 years after issuance relative to the issuance date for the total sample.
As we have shown in Section 1.3, behavioral theories can explain the return dynamics of seasoned equity issuers but cannot fully explain the peculiar risk pattern we observe in the graphical and tabular results. From a real options perspective however, these results make sense. Indeed the “risk loadings” should increase prior to issuance as the leverage of the growth option(s) held by the issuer rises. Risk should decrease after issuance when the option is unlevered through real investment.
An alternative explanation for the perceived risk dynamics is a mix of growth options and behavioral elements such as the “market timing” ability of astute managers. Such managers are able to optimally time the SEO and the subsequent investment when market conditions are good and/or the equity is overvalued such that existing shareholders will not see their holdings diluted. Conversely, a firm would issue debt when it is undervalued (Choe, Masulis, and Nanda 1993).
In a third scenario we may assume that all firms, as economic agents, are rational and pursue a well-defined goal when raising capital on the capital market. While for some of them it can be debt repayment or acquisition financing, a great number of firms aim to implement capital investments.
Prior to issuance, the market anticipates the uncertainty associated with the existence of the “window of investment opportunity” – the real options that a firm
26 is being exposed to, and reacts by increased volatility and correlation with the overall market conditions (indeed, the more favorable the overall economic situation is, the more chances for the new projects to succeed).
After the issuance, unsurprisingly, investors obtain more information about the future cash-flows of the projects financed and the uncertainty is cleared. Thus the risk decreases sharply and remains much lower afterwards.
3.1.4.2 Country sub-samples
To get a perception of the relative risk dynamics occurring on distinct capital markets, we performed the average beta analysis for each Nordic SEO subsample in addition to the aggregate sample. Since the number of firms in each country subsample is now reduced accordingly, we should be fairly cautious in drawing rigid conclusions or making straightforward risk dynamics comparisons.
We can observe a pronounced risk change around the issuance event for Norway and Sweden in line with the dynamics highlighted by Carlson et al. (2010): a perceived increase in risk prior to the SEO event and a smooth decrease thereafter.
However, the average risk dynamics of the Danish issuers is quite noisy, probably due to the very small sample or to the inaccuracy of the data (the electronic list of Danish issues is poorly informative about the actual types of issues reported).
The contemporaneous beta series obtained with monthly, semiannual and annual estimation windows can also be compared to each other (refer to Figure 2 in the Appendix). Definitely, the visual trends in average betas across the three countries look similar regardless of the estimation period, although the noise of the beta values decreases and the trend becomes more pronounced with the widening of the estimation window (from monthly to annual). The noise reduction induced by the annual estimation windows can account for the more significant differences in the annual pre- and post-issuance beta estimates reported in Table 2.
3.1.4.3 R&D sub-samples
The articles we are focusing on (e.g. Lyandres et al. (2008)) try to explain the dynamics of stock returns around issuance events solely based on real investment.
27 However, there seems to be a positive relationship between research and development activities in a firm and its stock returns, as documented by Chan, Lakonishok and Sougannis (2001). In particular they provide evidence that R&D intensity is positively associated with return volatility, ceteris paribus. In the real options terminology, Chan et al (2001) write that R&D actually generates risky expansion options, whereas only real investment transforms them into less risky assets in place.
Based on this intuition, we reorganized the subsamples of issuers based on research and development intensiveness6. The financial industries as well as unclassified firms were excluded from the classification.
Table 3 in the Appendix is based on the accounting item R&D-to-sales recorded in Datastream from 2003-2011 for listed Norwegian, Swedish, and Danish companies. We considered this relatively long time interval in order to capture a time-consistent, persistent dynamics rather than a momentary picture. To avoid outliers, we estimated the median annual expenditure on R&D for each company during this period, and after classifying each company according to its industry – we computed the median annual R&D-to-Sales for each industry. Finally, we assigned the industry medians to one of three categories: High, Medium or Low R&D-to-Sales based on the top 30%, middle 40% and bottom 30% deciles.
The distribution of the R&D intensiveness is generally as expected. It makes intuitive sense that industries such as Pharmaceuticals and Biotechnology, Aerospace and Defense and Software and Computer Services (tertiary economy) to be R&D intensive, and industries such as Mining and Forestry and Paper (primary economy) to be at the lower end of R&D intensiveness.
For this analysis, we used the original aggregate sample of equity issuers with the same initial constraints and in addition, with firms classified into the “H” (30%
high), the “M” (40% medium), or the “L” (30% low) category along the R&D-to- sales dimension. The beta dynamics by R&D intensive industry sub-samples displays vivid tendencies. Thus, compared to less R&D intensive industries, the
6 We have proxied the R&D intensiveness by the R&D-to-Sales ratio.
28 industries generating most of the real options by spending the highest amounts on R&D were subject to a sharper increase in risk before the issuance, and a sharper drop afterwards.
Numerical evidence on the beta estimates differences described in Table 10 from the Appendix also indicate the significant 2 and 3 year annual post-issue differences for the highly R&D – intensive industries. This again could suggest that real options realizations impact mostly real-option sensitive companies, namely those that by employing intensive research programs accumulated substantial uncertainty which resolved after the market financing and thus realization of these investment possibilities.
One might argue that the fall in beta could also be induced by a drop in financial leverage following the stock issue. Nonetheless, a simple leverage theory would predict a more abrupt beta decrease after the issue, and would not explain the pre- issuance beta run-up. In Section 3.1.6 we have regressed changes in betas (i.e. the difference between the post-issuance and the pre-issuance betas) on both a proxy for leverage change and a proxy for real investment to check the validity of a financial leverage-based explanation of the observed risk dynamics.
0 0,2 0,4 0,6 0,8 1 1,2
-520 -390 -260 -130 0 130 260 390 520 650 780
Average βeta value
Figure 4: Average semiannual beta classified wrt R&D intensivneness
Highly R&D intensive industries Medium R&D intensive industries Low R&D intensive industries
Days before (negative values) and after (positive values) issuance
29 3.1.5 The case of Blom ASA
The methodology and the results obtained based on the aggregate sample of all eligible SEOs may become clearer if we take the specific case of one firm from our sample. Blom ASA is a Norwegian geographical information and offshore technology company founded in 1954. Starting from its listing on Oslo Stock exchange in 1988 (OSE: BLO) Blom has been steadily expanding its business both by organic growth and mergers and acquisitions. Prior to the equity issuance of our interest, on 13 May 1997 news highlighting a merger proposal between Blom ASA and CreditInform ASA emerged. Board considered the necessity to increase share capital by up to NOK 4,000,000 by equity issue. “The reason is partly because the company may need to strengthen the equity in connection with efforts internationally and within the information systems and information technology and partly to be able to complete acquisitions and establishments of enterprises with settlement in shares and/or cash”7
The amount of shares increased four times in August 1997 from 2848 to 113928. Both investigation of OSE listings information and the DATASTREAM Number of shares datatype have revealed the absence of additional major seasoned equity offerings in the period 1995 – 2000 (the five year window of our study). Figure 5 in the Appendix vividly illustrates how the risk captured by the contemporaneous semiannual market beta grows steadily before the issuance event (going up for two years of growth without major downturns) and decreases abruptly during the next three years. Based on the real options explanation, we can infer that uncertainty associated with the firm expansion plans has pushed up the risk of the company. After issuance, which supposedly was followed by investment in new business units and technologies, active and potential investors could have gained more information about the intrinsic value of the company.
7Own translation of the citation from the Factiva news Document reutno0020011003dt5d0075w http://global.factiva.com/aa/?ref=reutno0020011003dt5d0075w&pp=1&fcpil=en&napc=S&sa_fro m=
8 Datastream data
30 3.1.6 Market beta, leverage and real options
Intuitively, the real options based explanation of the risk dynamics around equity offerings is tempting. Nonetheless, still intuition says that financial leverage can also account for the observed dynamics. Indeed, with an increase in equity, financial leverage is expected to decrease after issuance. This should lead to a decrease in systematic risk and to a subsequent drop in market beta. This section will empirically explore two potential determinants of systematic risk dynamics around SEO events: financial leverage and real investment.
The relationship between market beta and accounting measures of risk such as financial leverage is a traditional area of academic research. Most theoretical literature revealed a positive and linear relationship between required return and leverage, as formulated by Modigliani and Miller (1958, 1963). Modigliani and Miller show that the required rate of return on equity of a levered firm increases proportionally to the debt-to-equity ratio.
Hamada (1972), and later on Bowman (1980) independently designed two closely related linear models in order to capture the relationship between the beta of an unlevered firm and the beta of the same firm, if levered. The Bowman model, ignoring corporate income tax, is:
-Be = Bu (1 + D/E) = Bu + Bu D/E, where -Bu = Unlevered beta (asset beta);
-Be = Levered beta which equals beta of the common stock;
-D = Market value of debt of the levered firm;
-E = Market value of equity of the levered firm.
Many other studies have investigated the determinants of systematic risk as measured by market beta, focusing on financial characteristics such as operating risk, changes in financial leverage, size and liquidity. Reviewing 13 empirical studies of the determinants of risk, (Ang, Peterson, and Peterson 1985) conclude that there is great variation among the models in their specification and empirical results. Furthermore, several of these studies fail to provide clear justification or hypotheses for the role of particular variables in the models or for the specified functional form. Beside financial leverage, several authors have supported the use
